Microsoft expands Project Perception to monitor risky AI agents

Microsoft is expanding Project Perception to bring continuous oversight to autonomous AI agents. The platform is designed to discover agents operating across an environment, assign each a distinct identity, and attach a set of permissions that govern what the agent can access or do.

Once an agent is known to the system, Project Perception monitors its actions in real time. When an agent’s behavior requires intervention, designated guardian agents can step in directly. The approach moves the security model away from securing a static application and toward supervising digital actors that act independently over time.